checkpoint-1190/trainer_state.json

{
  "best_global_step": 1190,
  "best_metric": 0.2564544677734375,
  "best_model_checkpoint": "siglip2-finetune-full/checkpoint-1190",
  "epoch": 2.0,
  "eval_steps": 500,
  "global_step": 1190,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.8403361344537815,
      "grad_norm": 21.780553817749023,
      "learning_rate": 1.4812680115273776e-05,
      "loss": 0.7499,
      "step": 500
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8325094657130837,
      "eval_loss": 0.4362837076187134,
      "eval_model_preparation_time": 0.0022,
      "eval_runtime": 307.2598,
      "eval_samples_per_second": 61.889,
      "eval_steps_per_second": 7.736,
      "step": 595
    },
    {
      "epoch": 1.680672268907563,
      "grad_norm": 19.866661071777344,
      "learning_rate": 9.048991354466859e-06,
      "loss": 0.4428,
      "step": 1000
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.9009255363904081,
      "eval_loss": 0.2564544677734375,
      "eval_model_preparation_time": 0.0022,
      "eval_runtime": 304.8059,
      "eval_samples_per_second": 62.387,
      "eval_steps_per_second": 7.798,
      "step": 1190
    }
  ],
  "logging_steps": 500,
  "max_steps": 1785,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 3.1854682608775004e+18,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": null
}